some sockets are left open about engine.io

[rauchg]

@kapouer good find. It seems to me that some sockets are throwing instead of emitting error events, therefore never getting properly shut down.

What do you think about this assessment @einaros ?

[einaros]

In the last version of ws, much has been done to take care of this. I suspect that the websocket handlers in websocket.io, which are not part of the parsers imported from ws, could need similar measures.

Ideally, as much as possible of what websocket.io is doing should be left to ws, so that fixes can land in one place rather than several :)

On 8. juli 2012, at 23:07, Guillermo [email protected] wrote:

@kapouer good find. It seems to me that some sockets are throwing instead of emitting error events, therefore never getting properly shut down.

What do you think about this assessment @einaros ?

---

Reply to this email directly or view it on GitHub:
#33 (comment)

[rauchg]

Sure but it's ws:

at WebSocket.send (/xxx/node_modules/engine.io/node_modules/websocket.io/node_modules/ws/lib/WebSocket.js:175:16)

[einaros]

I'm not entirely up to speed on how websocket.io's hybi.js uses ws. Which version of ws is running here? As I said, much has been done in the last release. If an earlier one is in use, then I am not surprised. If it is the latest version, I will have to look into how the library is being used, and see how exactly this can occur.

@nicokaiser, have you seen any improvements with regards to lingering sockets using barebones ws lately?

[nicokaiser]

@einaros Currently it looks quite good, no crashes and only moderate growth in memory consumption (apart from Node not releasing unused memory). But, I set custom "error" and "timeout" handlers to the socket before handing them to ws.handleUpgrade. The wsio.Socket does the same, so there should be no difference.

[kapouer]

Since engine.io 0.1.2 it seems the errors i'm seeing have changed :

Error: reserved fields must be empty
at WebSocket.onError (/home/test/node_modules/engine.io/lib/transport.js:77:13)
at WebSocket.emit (events.js:70:17)
at Receiver.onerror (/home/test/node_modules/engine.io/node_modules/ws/lib/WebSocket.js:542:10)
at Receiver.error (/home/test/node_modules/engine.io/node_modules/ws/lib/Receiver.js:301:8)
at Receiver.processPacket (/home/test/node_modules/engine.io/node_modules/ws/lib/Receiver.js:187:10)
at Receiver.add (/home/test/node_modules/engine.io/node_modules/ws/lib/Receiver.js:93:24)
at Socket.firstHandler (/home/test/node_modules/engine.io/node_modules/ws/lib/WebSocket.js:500:22)
at Socket.emit (events.js:67:17)
at TCP.onread (net.js:367:14)

Note that the client is still 0.1.0 so kill me if that's the reason.

[rauchg]

/me kills him

On 0.1.2 we upgraded the ws transport. @einaros do you have any guesses what could be causing this?

[rauchg]

@kapouer do the sockets stay open in this case too?

[kapouer]

it's hard to mesure so i'm not sure, but i'm almost sure there are much less sockets left open. The ratio i mesure has been staying low for days.

[rauchg]

Just restart engine.io every 10 connections (reference) (just kidding)

[nicokaiser]

I'm also getting these errors with plain ws, but I thought this was because of broken clients or proxies.

In my case, the first WebSocket packet they sent was the header again, yet I was not able to reproduce this...

Am 04.08.2012 um 17:25 schrieb Guillermo Rauch [email protected]:

/me kills him

On 0.1.2 we upgraded the ws transport. @einaros do you have any guesses what could be causing this?

---

Reply to this email directly or view it on GitHub:
#33 (comment)

[einaros]

Interfering proxies or client issues may cause that. I'm pretty confident a
ws client setver combo without intermediates wouldn't cause that.
On Aug 4, 2012 5:53 PM, "Nico Kaiser" <
[email protected]>
wrote:

I'm also getting these errors with plain ws, but I thought this was
because of broken clients or proxies.

In my case, the first WebSocket packet they sent was the header again, yet
I was not able to reproduce this...

Am 04.08.2012 um 17:25 schrieb Guillermo Rauch [email protected]:

/me kills him

On 0.1.2 we upgraded the ws transport. @einaros do you have any guesses
what could be causing this?

---

Reply to this email directly or view it on GitHub:
#33 (comment)

---

Reply to this email directly or view it on GitHub:
#33 (comment)

[rauchg]

Thanks. Closing this for the time being.